by John Allen Paulos
With apologies to Charles Dickens, it will be the best of times, it will be the worst of times.
In his recent book, Deep Utopia: Life and Meaning in a Solved World, philosopher Nick Bostrom, the author of Superintelligence, speculates about human and trans-human lives after AI has developed to a kind of fearsome maturity at some indeterminate point in the future. What do our descendants do when AI can do virtually everything faster, more efficiently, better than they can? What, if anything, will be worth doing is the question underlying much of the book. Will we be become terminally jaded without purpose or will be become, as Bostrom puts it, mere hedonistic “pleasure blobs”?
The book provides a bountiful wealth of details, speculations, and extrapolations on boredom, interestingness, repetitiveness, and activities that might replace and make up for our loss of functional work and meaningful vocations. He mentions activities such as amateur art, music, personal reveries, social interactions, gardening, and the like.
The closest contemporary version of our distant descendants might be very wealthy young retirees or entitled trust fund kids. Still, the latter are not at all the purposeless self-indulgent, bored, unproductive, nihilistic descendants he first sketches. Gradually Bostrom attempts to complexify and significantly brighten this dim characterization of our distant future utilizing a variety of abstract philosophical arguments citing Malthus, Nozick, Thaddeus Metz, and many others. (The conceit is that these lectures are delivered by a professor (essentially Bostrom) to three students who rarely ask questions.)
On this more upbeat note Bostrom argues that our descendants might devise pursuits that, though vastly better performed by AI, could still interest them if human constraints were placed on these pursuits. And unusual physical feats, fanciful games with arbitrary rules, and the like might provide a kind of ersatz purpose for them. There is a subtle, almost Talmudic discussion of what constitute interestingness and how to prolong it, especially if our future life spans will be so much longer.
With regard to hedonism, he makes clear he’s talking not only about sexual promiscuity but about physical and mental pleasures generally. He also examines at length the relationship of pleasure to other long-term values, to ethical considerations, and to issues of fulfillment, which issue is also meticulously analyzed.
Hedonism and interestingness suggest the question of what might constitute new and exciting breakthroughs and experiences. Bostrom asks if there is anything new and exciting that can match being born, learning that other minds exist, discovering locomotion, and a few other early life experiences.
Of course, nothing compares with these, but what if we were capable of changing not only our physical looks and attributes, but also our psychologies and memories via AI-mediated developments in biology and neuroscience. He writes that such drastic and ultra-rapid alterations of our internal selves would be possible with future and futuristic AI, but warns they would seriously weaken our sense of personal identity which is constructed in large part though the continuity of our memories. This latter point and a few others in the book remind me of some related issues discussed in Derek Parfit’s Persons and Reasons.
If AI will give our descendants the capability to make all these amazing changes to their cognition, memory, psychology, and physical selves, then who really will they be? Will they even be the same species as us or will they be trans-humans or post-humans? The natural secular interpretation of the biblical verse comes to mind: ”For what is a man profited if he shall gain the whole world, and lose his own soul.”
There is in the book a plaintive whisper of the question, what’s it all about, what’s the point. Or maybe I’m just projecting. Is life just more and more and more of the same, which brings me to this marginally relevant post script:
Let me describe a mathematical way to illustrate a significant aspect of boredom and jadedness that I’ve mentioned elsewhere. So blame me and not Bostrom for this illustration. Let’s imagine some random quantity, say the number of heads in 1,000 flips of a coin. (1,000,000,000 would be better, but let’s try to avoid a virtual carpal tunnel pain in our wrists.) Imagine further that after flipping a coin 1,000 times and counting the number of heads, you again flip the coin 1,000 times and count the number of heads. If you do this a third time and a fourth time and keep on flipping the coin 1,000 times and counting the number of heads, how often will you achieve a record number of heads; that is, how often will you achieve a number of heads greater than the number of heads in all previous instances of 1,000 flips?
To further illustrate, let’s assume that the number of heads the first ten times you flip the coin 1,000 times is 524; 496; 501; 527; 488; 499; 514; 519; 474; 531. The first result of 524 will, of course, automatically be a record number of heads, but the next record number of heads—527—occurs the fourth time you flip the coin 1,000 times, and the next record of heads after that, 531—the third record number, occurs on the tenth time you flip the coin 1,000 times. That is, the sequence or records R and un-records U would in this case be RUURUUUUUR. Note that the records occur less and less frequently. Moreover, you probably wouldn’t achieve much more than 9 Rs, 9 record numbers of heads, by the ten thousandth time you flipped the coin 1,000 times, and even by the one millionth time you flipped the coin 1,000 times you probably wouldn’t achieve much more than 14 Rs, 14 record numbers of heads. Amazingly, but not coincidentally, the 9th root of 10,000 and the 14th root of 1,000,000 are approximately equal to e, which is 2.71828 . . . , the base of the natural logarithm. This is one reason I like this thought experiment.
Summarizing, I can say that the general relation is that if the number of times you flip a coin 1,000 times is a number N (a hundred, a million, a billion, whatever), then the number of records—that is, the number of Rs you will achieve—will be roughly equal to the natural log of N, and this is a shrinking fraction of N. And the relevance to the loss of excitement and verve that aging usually brings about? Interpreting record highs loosely along whatever dimension you care to mention, the mathematical fact just mentioned suggests that these high points—records, peak experiences, and the like—will occur less and less frequently as we get older
You might say that the above argument depends on events occurring randomly, as the results of the 1,000 coin flips did, and that’s true. The establishment of new records in sports, for example, occurs more frequently than that because of improvements in training and nutrition. Still, most common discoveries, epiphanies, achievements (records for short) do tend to occur early rather than late.
There’s seems to be no good way to get around eventual boredom and jadedness unless we can somehow manage to start over, the possibility of which Bostrom doesn’t shrink from discussing. In fact, he doesn’t shrink from discussing many such vexing issues. Oddly, given its length, the book ends abruptly and probably wisely without any focused remarks on the meaning of life in a post-AGI world. This isn’t surprising since he earlier in the book writes. “Asking someone the meaning of life is like asking their recommendation for shoe size.”
Deep Utopia is challenging, well-reasoned, and fun to read.
***
John Allen Paulos is an emeritus Professor of Mathematics at Temple University and the author of Innumeracy and A Mathematician Reads the Newspaper. These and his other books are available here.